Infrared image flat field noise correction method and device based on noise model

Abstract

The invention discloses an infrared image flat field noise correction method and device based on a noise model. The method comprises the following steps: firstly, acquiring an infrared original image Ii at different temperatures, and extracting Gaussian noise of the infrared original image Ii as an initial Gaussian noise template at the temperature; performing difference to obtain initial Gaussian noise templates under all temperatures in the working environment temperature range; then according to the working environment temperature of the real-time infrared image, parameters of the initial Gaussian noise template at the working environment temperature are finely adjusted, and the fine adjustment Gaussian noise template corresponding to the minimum distance between the Gaussian noise template after the fine adjustment parameters are extracted and the real-time infrared image is the updated Gaussian noise template at the working environment temperature; and finally, subtracting the updated Gaussian noise template of the Gaussian noise template updating module from the real-time infrared image to realize noise reduction of the real-time infrared image. According to the method, real-time updating can be carried out to ensure real-time correction of the infrared image, and the stability of the image is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of uncooled infrared imaging, and in particular to a method and device for correcting flat-field noise of an infrared image based on a noise model. Background Art

[0002] Uncooled infrared imaging sensors have uneven brightness due to process limitations, uneven sensor response, illumination differences, or other reasons, which manifests as inconsistent responses between the center and edges of the image. Currently, the most common methods for dealing with uneven responses are calibration correction or noise estimation. The calibration correction method uses a single or two-point response to correct the image by measuring the radiation response of a uniformly radiated blackbody under high and low temperature conditions. This method is based on the premise that the detector response has temporal stability. However, in real-world scenarios, rapid changes in ambient temperature and differences in detector temperature can easily cause inaccurate calibration correction or inability to adapt to environmental changes. The noise estimation method relies on the estimation ability of the algorithm, which can easily cause problems such as overestimation and estimation failure during the estimation process, resulting in missing image details or the introduction of new noise. How to update and correct the noise model in real time is very important for maintaining long-term stability of infrared. Summary of the invention

[0003] In view of this, the present invention provides a method and device for flat-field noise correction of infrared images based on a noise model, which determines the parameters of the noise model through initial calibration, and updates the parameters in real time when a real-time image is acquired, thereby realizing flat-field correction of the infrared image.

[0004] The infrared image flat field noise correction method based on the noise model of the present invention comprises the following steps:

[0005] Step 1: Obtain the original infrared images I at different temperatures (assuming N images are collected) i , i = 1, 2, ..., N. The acquisition method of each image is as follows:

[0006] Assume that the temperatures of N temperature points are T i , i = 1, 2, ..., N, then by placing the imaging components in a temperature environment of T i The original image I can be obtained by placing the photosensitive surface of the imaging component in a constant temperature box and facing the black body with a uniform temperature field. i .

[0007] Step 2: Obtain the original infrared image I i Extraction temperature T i The initial Gaussian noise template under the condition is stored, where the Gaussian model template is:

[0008]

[0009] a i ,b i ,c i ,d i ,e i ,f i is the Gaussian model fitting parameter to be solved, which can be obtained by minimizing the fitting error function E(a) that measures the difference between the fitting surface and the actual data points. i ,b i ,c i ,d i ,e i ,f i ) is obtained, and the specific fitting error function formula is as follows:

[0010]

[0011] Where W and H are the width and height of the image.

[0012] Step 3: Obtain the initial Gaussian noise template parameters under specific temperature conditions T by interpolation, a T ,b T ,c T ,d T ,e T ,f T The calculation method is:

[0013] a T =L(a i )

[0014] b T =L(b i )

[0015] c T =L(c i )

[0016] d T =L(d i )

[0017] e T =L(e i )

[0018] f T =L(f i )

[0019] Where L is the interpolation function, which can be linear interpolation, quadratic function interpolation, or polynomial interpolation.

[0020] Step 4: According to the real-time infrared image and the initial Gaussian noise template parameter a under the specific temperature condition T T ,b T ,cT ,d T ,e T ,f T The Gaussian noise template is updated in real time within a certain threshold range. The updated Gaussian noise template is:

[0021]

[0022] Updated Gaussian noise template parameter a T ′ ,b T ′ ,c T ′ ,d T ′ ,e T ′ ,f T ′ By calculating the minimum Euclidean distance D within the range of Δa, Δb, Δc, Δd, Δe, Δf, we can get:

[0023]

[0024] Step 5: Obtain the real-time image after noise reduction, and the calculation method is as follows:

[0025] I out =I T -G

[0026] I out For image I T Final image after denoising.

[0027] The present invention also provides an infrared image flat field noise correction device based on a noise model, comprising: a memory, a Gaussian noise template extraction module, a Gaussian noise template update module and a denoising module; the above method is used for denoising; wherein,

[0028] The memory stores the temperature T of different sampling points within the working environment temperature range of the uncooled infrared imaging sensor. i The infrared original image I i ;

[0029] The Gaussian noise template is used to extract the temperature T of each sampling point i The infrared original image I i The Gaussian noise of is used as the initial Gaussian noise template at the temperature, and the initial Gaussian noise template at all temperatures within the working environment temperature range is obtained by difference;

[0030] The Gaussian noise template updating module fine-tunes the parameters of the initial Gaussian noise template at the working environment temperature according to the working environment temperature of the real-time infrared image, calculates the distance between the Gaussian noise template after each fine-tuning parameter and the real-time infrared image, and extracts the Gaussian noise template corresponding to the minimum distance as the updated Gaussian noise template at the working environment temperature;

[0031] The denoising module subtracts the updated Gaussian noise template of the Gaussian noise template updating module from the real-time infrared image to achieve denoising of the real-time infrared image.

[0032] The present invention also provides a computer program product for infrared image flat field noise correction based on a noise model, the computer program product comprising a non-volatile readable storage medium and a computer program, the computer program being tangibly stored on the non-volatile readable storage medium, and the computer program being executed by a processor in an FPGA, DSP, ARM or other computer to implement the steps of the above correction method.

[0033] Beneficial effects:

[0034] 1) Determining the parameters of the noise model through initial calibration and updating the parameters in real time when real-time images are acquired can ensure real-time correction of infrared images and ensure image stability.

[0035] 2) The algorithm complexity of this technology is low, and it is easy to implement on different platforms such as FPGA, DSP, ARM, and PC. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The figure is a flow chart of the method of the present invention.

[0037] Figure 2 The comparison diagram of the effect before and after treatment by the method of the present invention is shown in Figure 1. a) is the picture before treatment, and b) is the picture after treatment. Figure 2 It can be seen that the image processed by the patented method has an obvious correction effect on flat field noise compared to the original image. DETAILED DESCRIPTION

[0038] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0039] The present invention provides a method for correcting flat-field noise of infrared images based on a noise model, taking an infrared core component with an operating temperature range of -40°C to 60°C and a resolution of 640×512 as an example:

[0040] Step 1: Obtain and collect infrared original images I at different temperatures (assuming 6 images are collected) i , i = 1, 2, ..., 6. The acquisition method of each image is as follows:

[0041] The temperature range of the acquisition is -40℃ to 60℃, the temperatures of the 6 temperature points are -40℃, -20℃, 0℃, 20℃, 40℃, 60℃, and the temperature acquisition step is 20℃. The original image I can be obtained by placing the imaging component in a constant temperature box at the 6 temperature points of the temperature environment in turn and facing the photosensitive surface of the imaging component to the black body with a uniform temperature field. i .

[0042] Step 2: Obtain the original infrared image I i The initial Gaussian noise templates under 6 temperature conditions are extracted and stored, where the Gaussian model template is:

[0043]

[0044] a i ,b i ,c i ,d i ,e i ,f i is the Gaussian model fitting parameter to be solved, which can be obtained by minimizing the fitting error function E(a) that measures the difference between the fitting surface and the actual data points. i ,b i ,c i ,d i ,e i ,f i ) is obtained, and the specific fitting error function formula is as follows:

[0045]

[0046] Step 3: Obtain the initial Gaussian noise template parameters at any temperature T within the working temperature range by interpolation. Taking linear interpolation as an example, a T ,b T ,c T ,d T ,e T ,f T The calculation method is:

[0047]

[0048] Where T belongs to T i ~T i+1 within the range.

[0049] Step 4: Extract the initial Gaussian noise template parameter a under the working environment temperature condition T corresponding to the real-time infrared image T ,b T ,c T ,d T ,e T ,f T, fine-tune the template parameters within a certain threshold range and update the Gaussian noise template; the updated Gaussian noise template is:

[0050]

[0051] Updated Gaussian noise template parameter a T ′ ,b T ′ ,c T ′ ,d T ′ ,e T ′ ,f T ′ By calculating the minimum Euclidean distance D within the set threshold Δa, Δb, Δc, Δd, Δe, Δf, we can get:

[0052]

[0053] Step 5: Obtain the real-time image after noise reduction, and the calculation method is:

[0054] I out =I T -G

[0055] I out For image I T Final image after denoising.

[0056] The present invention also provides an infrared image flat field noise correction device based on a noise model, comprising: a memory, a Gaussian noise template extraction module, a Gaussian noise template update module and a denoising module; the above method is used for denoising; wherein,

[0057] The memory stores the temperature T of different sampling points within the working environment temperature range of the uncooled infrared imaging sensor. i The infrared original image I i ;

[0058] The Gaussian noise template is used to extract the temperature T of each sampling point i The infrared original image I i The Gaussian noise of is used as the initial Gaussian noise template at the temperature, and the initial Gaussian noise template at all temperatures within the working environment temperature range is obtained by difference;

[0059] The Gaussian noise template updating module fine-tunes the parameters of the initial Gaussian noise template at the working environment temperature according to the working environment temperature of the real-time infrared image, calculates the distance between the Gaussian noise template after each fine-tuning parameter and the real-time infrared image, and extracts the Gaussian noise template corresponding to the minimum distance as the updated Gaussian noise template at the working environment temperature;

[0060] The denoising module subtracts the updated Gaussian noise template of the Gaussian noise template updating module from the real-time infrared image to achieve denoising of the real-time infrared image.

[0061] The present invention also provides a computer program product for infrared image flat field noise correction based on a noise model, the computer program product comprising a non-volatile readable storage medium and a computer program, the computer program being tangibly stored on the non-volatile readable storage medium, and the computer program being executed by a processor in an FPGA, DSP, ARM or other computer to implement the steps of the above correction method.

[0062] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for correcting flat-field noise of infrared images based on a noise model, characterized in that: include: Step 1: Obtain the temperature T of different sampling points within the working environment temperature range of the uncooled infrared imaging sensor i The infrared original image I i , i=1,2,…,N; Step 2: Assuming the noise is Gaussian noise, extract the temperature T of each sampling point i Lower infrared original image I i Gaussian noise, as the temperature T i The initial Gaussian noise template; Step 3, interpolating the parameters of the initial Gaussian noise template at each sampling point temperature obtained in step 2 to obtain the initial Gaussian noise template at all temperatures within the working environment temperature range; Step 4: Update the initial Gaussian noise template under the working environment temperature according to the working environment temperature of the real-time infrared image, specifically: Fine-tune the parameters of the initial Gaussian noise template at the working environment temperature, calculate the distance between the Gaussian noise template after each fine-tuning parameter and the real-time infrared image, and the Gaussian noise template corresponding to the minimum distance value is the updated Gaussian noise template; Step 5: Subtract the Gaussian noise template updated in step 4 from the real-time infrared image to achieve noise reduction of the real-time infrared image.

2. The method according to claim 1, characterized in that In step 1, the imaging components are placed in a temperature environment of T i The original image I is obtained by placing the photosensitive surface of the imaging component facing the black body with a uniform temperature field. i .

3. The method according to claim 1, characterized in that In the step 2, the expression of Gaussian noise is: Where (x, y) is the pixel coordinate of the original infrared image; a i ,b i ,c i ,d i ,e i ,f i is the Gaussian model fitting parameter to be solved, and the fitting error function E(a) that measures the difference between the Gaussian fitting surface and the actual data points is minimized. i ,b i ,c i ,d i ,e i ,f i ) is obtained; the fitting error function is: Where W and H are the width and height of the image respectively.

4. The method according to claim 3, characterized in that In step 4, the updated Gaussian noise template is: Updated Gaussian noise template parameter a T ′ ,b T ′ ,c T ′ ,d T ′ ,e T ′ ,f T ′ By calculating the minimum Euclidean distance D within the set range of Δa, Δb, Δc, Δd, Δe, Δf, we can get:

5. The method according to any one of claims 1 to 4, characterized in that: In the step three, the interpolation is linear interpolation, quadratic function interpolation or polynomial interpolation.

6. An infrared image flat field noise correction device based on a noise model, characterized in that: include: A memory, a Gaussian noise template extraction module, a Gaussian noise template update module and a denoising module; denoising is performed using the method described in claims 1 to 5; wherein, The memory stores the temperature T of different sampling points within the working environment temperature range of the uncooled infrared imaging sensor. i The infrared original image I i ; The Gaussian noise template is used to extract the temperature T of each sampling point i The infrared original image I i The Gaussian noise of is used as the initial Gaussian noise template at the temperature, and the initial Gaussian noise template at all temperatures within the working environment temperature range is obtained by difference; The Gaussian noise template updating module fine-tunes the parameters of the initial Gaussian noise template at the working environment temperature according to the working environment temperature of the real-time infrared image, calculates the distance between the Gaussian noise template after each fine-tuning parameter and the real-time infrared image, and extracts the Gaussian noise template corresponding to the minimum distance as the updated Gaussian noise template at the working environment temperature; The denoising module subtracts the updated Gaussian noise template of the Gaussian noise template updating module from the real-time infrared image to achieve denoising of the real-time infrared image.

7. A computer program product for infrared image flat field noise correction based on a noise model, characterized in that: It comprises a non-transitory readable storage medium and a computer program, wherein the computer program is tangibly stored on the non-transitory readable storage medium, and the computer program is executed by a processor in an FPGA, DSP, ARM or other computer to implement the steps of the correction method as described in any one of claims 1 to 5.

Images